摘要： 通过评估并预测当前手机通信网络中移动用户之间的通话特征,可使运营商掌握消费者行为规律。为此,提出一种新的手机通信网络链路预测模型。选取共同邻居作为手机通话网络中用户对的基本相似性指标,同时考虑用户通话时长、通话时长比等关键特征的差异,设计节点相似性经验公式,并基于部分真实手机通话数据获得经验公式中的3个关键参数。应用于某通信运营商50 000个手机用户数据集的链路预测结果表明,该预测模型能准确分析手机用户的消费行为模式。

关键词: 节点相似性, 手机通信网络, 链路预测, 社会网络, 手机用户

Abstract: Evaluating and predicting calling characters among users in current mobile phone communication network enables mobile network operators to effectively obtain behavior rules of consumers.This paper proposes a new link prediction model for mobile phone communication network.Given common neighbors as the basic similarity metric among users in mobile phone communication network and the differences of some key calling characters such as the call duration and call duration ratio,a new node similarity experienced formula is proposed and three key parameters based on part of real mobile phone communication dataset are also obtained.The new node similarity metric is applied to the link prediction in a dataset generated by 50 000 mobile phone users in a communication operator.The results obtained from performance analysis show that the poropsed node similarity model can accurately analyze the consuming behavior patterns of mobile phone users.

Key words: node similarity, mobile phone communication network, link prediction, social network, mobile user

中图分类号: 

• TP393